[Master Index] [Index for export/lyngby]



Function Synopsis

[h] = lyngby_fir_main(x, Y, arg1, arg2, arg3, arg4, arg5, ...

Help text

 lyngby_fir_main      - Regularized FIR filter, main function 

       function H = lyngby_fir_main(x, Y, 'PropertyName',

       Input:    x   Input signal to the system - the paradigm (vector)
                 Y   Output signal of the system (vector or matrix). 

       Property: FilterOrder      {7} The order of the estimated FIR
                                  filter (integer), ie, filter length 
                 RegMethod        [ {Ridge} | RegInverse | PCR | SVD
                                  | Smooth ] Ridge regression
                                  (ridge/reginverse) or principal
                                  component regression (PCR/SVD) of
                                  smooth FIR 
                 PCRComp          Number of principal components for
                                  the PCR 
                 Regularization   {0} Regularization parameter for
                                  ridge regularization 
                 ConvType         [ {Skip} | Zeropad | Extend | Wrap ]
                                  Convolution type. if 'Zeropad' use
                                  all data points in the filter
                                  estimation. if 'Skip' it will
                                  disregard the FilterOrder-1 first
                                  data points. 'Extend' will use the
                                  first value of 'x': x(t) = x(1), t<1

       Output:   H   Estimated response function

       Estimation of the finite impulse response (FIR) filter, ie, a
       linear model with 'stick' functions. This is also a "one-layer
       feedforward linear neural network with weight decay" and an
       "ARX(0,n) model (with ridge regression)" 

       There are several regularization methods control with the
       'RegMethod' property:

       With the 'RegMethod' as 'Ridge' the regularization will
       be ridge regression, and the 'Regularization' parameter will
       be the ridge parameter. 

       The 'PCR' regularization method will first preprocess the input
       making basis function out of the input and use this in the
       regression. The regularization parameter will then function
       as a threshold parameter cutting off SVD component with a
       singular value lower than this threshold.

       Ref: Goutte et al., (2000), IEEE TMI, 19(12):1188+.


 $Id: lyngby_fir_main.m,v 1.18 2003/02/21 13:39:52 fnielsen Exp $

Cross-Reference Information

This function calls
This function is called by

Produced by mat2html on Wed Jul 29 15:43:40 2009
Cross-Directory links are: OFF